[mira_talk] Re: 35M Solexa paired-end reads on a Linux cluster with 128Gb RAM -> out of memory :o(
- From: Bastien Chevreux <bach@xxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Thu, 18 Mar 2010 21:25:01 +0100
On Donnerstag 18 März 2010 Martin A. Hansen wrote:
> Can I do anything about this?
> [...]
> Out of memory detected, exception message is: std::bad_alloc
Yes and no.
Strange, there shouldn't be an OOM at this stage as the memory has been pre-
allocated. But looking at the code I saw a faint possibility for memory re-
allocation of double the size of the current buffer, which in your case would
be 180 GiB. I quickly made a fix to this (and added some debugging stuff in
case this fails).
Can you please download
http://www.chevreux.org/tmp/mira_3.0.3x1_prod_linux-gnu_x86_64_static.tar.bz2
and see whether the bug is still there (or maybe a new error message pops up)?
In any case: even if the 'fix' isn't fixing the problem, you should be able to
circumvent that particular problem by limiting the process size to something
like 50 GiB with
-GE:mps=50
Regards,
Bastien
PS: Lionel is right: I currently cannot recommend Solexa only for high
coverages.
--
You have received this mail because you are subscribed to the mira_talk mailing
list. For information on how to subscribe or unsubscribe, please visit
http://www.chevreux.org/mira_mailinglists.html
Other related posts:
